HDAC inhibition imparts beneficial transgenerationaleffects in Huntington’s disease mice via altered DNAand histone methylationHaiqun Jiaa, Charles D. Morrisa, Roy M. Williamsb,1, Jeanne F. Loringb, and Elizabeth A. Thomasa,2

Departments of aMolecular and Cellular Neuroscience and bChemical Physiology, The Scripps Research Institute, La Jolla, CA 92037

Edited by Solomon H. Snyder, Johns Hopkins University School of Medicine, Baltimore, MD, and approved November 24, 2014 (received for review August21, 2014)

Increasing evidence has demonstrated that epigenetic factors canprofoundly influence gene expression and, in turn, influenceresistance or susceptibility to disease. Epigenetic drugs, such ashistone deacetylase (HDAC) inhibitors, are finding their way intoclinical practice, although their exact mechanisms of action areunclear. To identify mechanisms associated with HDAC inhibition,we performed microarray analysis on brain and muscle samplestreated with the HDAC1/3-targeting inhibitor, HDACi 4b. Path-ways analyses of microarray datasets implicate DNA methylationas significantly associated with HDAC inhibition. Further assess-ment of DNA methylation changes elicited by HDACi 4b in humanfibroblasts from normal controls and patients with Huntington’sdisease (HD) using the Infinium HumanMethylation450 BeadChiprevealed a limited, but overlapping, subset of methylated CpGsites that were altered by HDAC inhibition in both normal andHD cells. Among the altered loci of Y chromosome-linked genes,KDM5D, which encodes Lys (K)-specific demethylase 5D, showedincreased methylation at several CpG sites in both normal and HDcells, as well as in DNA isolated from sperm from drug-treatedmale mice. Further, we demonstrate that first filial generation(F1) offspring from drug-treated male HD transgenic mice showsignificantly improved HD disease phenotypes compared with F1offspring from vehicle-treated male HD transgenic mice, in associ-ation with increased Kdm5d expression, and decreased histone H3Lys4 (K4) (H3K4) methylation in the CNS of male offspring. Addi-tionally, we show that overexpression of Kdm5d in mutant HDstriatal cells significantly improves metabolic deficits. These find-ings indicate that HDAC inhibitors can elicit transgenerationaleffects, via cross-talk between different epigenetic mechanisms, tohave an impact on disease phenotypes in a beneficial manner.

epigenetic | transgenerational | histone | neurodegenerative | therapeutic

Epigenetic alterations and transcriptional dysregulation haveemerged as common pathological mechanisms in many neu-

rological disorders (1, 2). Accordingly, therapeutic approaches thattarget gene expression represent an encouraging new avenue ofexploration for these disorders. Numerous phase I and II clinicaltrials using histone deacetylase (HDAC) inhibitors are ongoing.Novel HDAC inhibitors with improved safety, brain penetration,potency, and selectivity have greatly advanced the therapeutic po-tential of these agents for a wide range of noncancer indications,including neurodegenerative disease, psychiatric disorders, addic-tion, and inflammatory disease states (3–6).Class I-specific, benzamide-type HDAC inhibitors have been

developed as a therapeutic approach for Friedreich’s ataxia (7, 8),and we have previously tested these inhibitors for their potentialbenefit in Huntington’s disease (HD). Our studies have shownthat inhibitors selectively targeting HDAC1 and HDAC3 arebeneficial in several different HD model systems (9–11). In par-ticular, in vivo studies have shown that pharmacological inhibitionof HDAC1 and HDAC3 led to improved disease-associated bodyweight loss, motor dysfunction, and cognitive decline in two dif-ferent HD mouse models (9, 11). These findings are consistent

with other studies showing beneficial effects of broadly actingHDAC inhibitors in HD mouse models (12–14). However, themechanisms by which these compounds are imparting benefit indisease models are unclear.In general, HDAC inhibitors are known to elevate acetylation

of histones, both globally and at specific loci, resulting in a morerelaxed chromatin structure that facilitates gene transcription.However, several microarray studies have demonstrated thatHDAC inhibitors can cause both up- and down-regulation of geneexpression patterns (11, 15), suggesting that HDAC inhibition mayalter the expression of other regulatory enzymes and/or cofactors,which subsequently act as activators or repressors of gene activity.Accumulating evidence suggests that many HDACs, includingclass I HDACs, can also deacetylate nonhistone proteins (16),which may contribute to their beneficial properties.In this study, we used gene expression microarrays to identify

potential mechanisms of action of the HDAC1/3-targeting in-hibitor, HDACi 4b. We find that HDAC inhibition alters theexpression of several DNAmethylation-related genes and, further,that HDAC inhibition could induce DNA methylation changes ona genome-wide scale. The potential for HDAC inhibitors to elicitalterations in DNA methylation is highly relevant because DNAmethylation is an epigenetic modification that can be inherited in

Significance

We demonstrate that histone deacetylase (HDAC) inhibitioncan elicit changes in DNA methylation in Huntington’s disease(HD) human fibroblasts, as well as in sperm from HD trans-genic mice, in association with DNA methylation-related geneexpression changes. We suggest that alterations in sperm DNAmethylation lead to transgenerational effects, and, accord-ingly, we show that first filial generation (F1) offspring ofHDAC inhibitor-treated male HD transgenic mice show im-proved HD disease phenotypes compared with F1 offspringfrom vehicle-treated male HD transgenic mice. These findingshave significant implications for human health because theyenforce the concept that ancestral drug exposure may bea major molecular factor that can affect disease phenotypes,yet in a positive manner. Further, we implicate Lys (K)-specificdemethylase 5d expression in this phenomenon.

Author contributions: H.J. and E.A.T. designed research; H.J., C.D.M., R.M.W., and E.A.T.performed research; J.F.L. contributed new reagents/analytic tools; H.J. and E.A.T. ana-lyzed data; and E.A.T. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The data reported in this paper have been deposited in the Gene Ex-pression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. GSE56963).1Present address: Institute of Genomic Medicine, Department of Pediatrics, University ofCalifornia, San Diego, La Jolla, CA 92093.

2To whom correspondence should be addressed. Email: bthomas@scripps.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1415195112/-/DCSupplemental.

E56–E64 | PNAS | Published online December 22, 2014 www.pnas.org/cgi/doi/10.1073/pnas.1415195112

the germ line and can lead to transgenerational effects in sub-sequent progeny; accordingly, we find that HD transgenic offspringof HDACi 4b-treated male HD transgenic mice exhibit improveddisease phenotypes compared with HD transgenic offspring fromvehicle-treated male HD transgenic mice. In addition, we dem-onstrate a role for the gene lysine (K)-specific demethylase 5d(Kdm5d) in this phenomenon, further implicating cross-talk amongdifferent epigenetic mechanisms. Identifying the relationship(s)between chromatin and DNA methylation is imperative for un-derstanding the programming of gene expression within the epi-genome and mechanisms of epigenetic transgenerational effects, aswell as for revealing the potential impact of the clinical use ofHDAC inhibitors.

ResultsWe have previously shown that HDAC1/3-targeting HDACinhibitors show therapeutic efficacy in HD mouse models, in as-sociation with elevated histone acetylation levels (9–11). Toidentify gene expression changes associated with altered histoneacetylation levels, we performed microarray analysis on skeletalmuscle samples from N171-82Q HD transgenic mice treated withthe HDAC1/3-targeting compound HDACi 4b and reanalyzedmicroarray data from brain samples of HDACi 4b-treated R6/2transgenic mice from our earlier studies (11). We used skeletalmuscle in light of the fact that many reports have implicatedskeletal muscle dysfunction as an important factor in motor dys-function, both in human and animal HD models (17–20). Usingtwo pathway analysis programs, Ingenuity Systems Pathway Anal-ysis (Ingenuity Systems) and Database for Annotation, Visualiza-tion, and Integrated Discovery (DAVID), we found that geneexpression alterations in response to HDACi 4b treatment weresignificantly associated with DNA methylation and chromatin or-ganization (Tables S1 and S2). In fact, the category “DNAmethylation and transcriptional repression signaling” was the thirdmost significantly associated canonical pathway related to drugtreatment, with 15 genes showing altered expression related to thispathway (Table S1). These 15 genes include two genes encoding

major DNA methyltransferase enzymes: DNA methyltransferase1 (Dnmt1) and DNA methyltransferase 3a (Dnmt3a); severalgenes encoding putative DNA demethylase enzymes: methyl-CpG binding domain proteins 3–5 (Mbds3–5), poly (ADP ribose)polymerase 1 (Parp1), and growth arrest and DNA damage-inducible 45 beta (Gadd45b) (21–24); and genes encoding pro-teins comprising two well-characterized class I HDAC-containingcorepressor complexes: the nucleosome remodeling and histonedeacetylase (NuRD) and Sin3 complexes (25). The fact that somany genes encoding proteins in these complexes were altered byHDACi 4b treatment strongly suggests a relationship betweenHDAC inhibition and alterations in DNA methylation andchromatin structure. The changes induced by HDACi 4b includeboth up-regulated and down-regulated genes, suggesting anoverall multifaceted regulation of these complexes. We validatedthe expression differences for some of these genes in muscle andbrain samples from HDACi 4b-treated WT and N171-82Qtransgenic mice using real-time PCR analysis (Table 1). Thegenes most significantly altered in expression were those encodingenzymes related to the regulation of DNA methylation: Dnmt1,Dnm3a,Gadd45b, Parp1, and RING finger protein 4 (Rnf4) (Table1). Increased expression of these genes could lead to either in-creased or decreased methylation of DNA.To test the direct effects of HDAC inhibition on DNA methyl-

ation genome-wide, we measured the locus-specific DNA methyl-ation patterns in cultured human fibroblasts from normal controlsand patients with HD using the Infinium HumanMethylation450BeadChip (Illumina, Inc.), which interrogates DNAmethylation at>450,000 CpG sites associated with both coding and noncodinggenes. First, comparing the HD vs. control condition, we found16,431 positions increased in methylation in HD cells relative tocontrol cells and 8,118 positions decreased in methylation, usinga multiple hypothesis corrected P value of 0.001 and a difference inmethylation cutoff of beta change of >j0.2j. These findings aresimilar to recently published work on striatal cells carryingpolyglutamine-expanded HTT (STHdh) that showed extensiveDNA methylation changes in mutant STHdhQ111 cells compared

Table 1. qPCR validation of DNA methylation-related genes altered by HDACi 4b treatment in cortex, striatum, and muscle

Symbol Entrez gene name

Cortex Striatum Muscle

Fold change P value Fold change P value Fold change P value

N171-82Q transgenic miceDnmt1 DNA (cytosine-5-)-methyltransferase 1 0.91 0.140 0.90 0.143 1.61* 0.025Dnmt3a DNA (cytosine-5-)-methyltransferase 3a 0.94 0.330 1.40* 0.032 1.23* 0.011Gadd45b Growth arrest and DNA-damage–inducible 45 beta 1.44* 0.049 1.01 0.460 0.96 0.338Hdac1 Histone deacetylase 1 0.81* 0.040 0.57* 0.011 0.81* 0.049Hdac2 Histone deacetylase 2 0.80 0.210 0.97 0.441 1.11 0.160Hdac3 Histone deacetylase 3 0.88 0.310 0.44* 0.031 0.98 0.437Mbd3 Methyl-CpG binding domain protein 3 1.57** 0.006 0.90 0.135 1.11 0.110Mecp2 Methyl CpG binding protein 2 0.97 0.890 1.05 0.360 1.21* 0.047Parp1 Poly (ADP ribose) polymerase family member 1 1.41* 0.007 1.22 0.107 0.92 0.186Rnf4 RING finger protein 4 1.11* 0.030 1.24* 0.035 0.87* 0.049

WT miceDnmt1 DNA (cytosine-5-)-methyltransferase 1 1.00 0.48 0.98 0.43 0.79 0.13Dnmt3a DNA (cytosine-5-)-methyltransferase 3a 0.78 0.16 1.37* 0.03 0.90 0.21Gadd45b Growth arrest and DNA-damage–inducible 45 beta 1.11 0.12 1.32* 0.03 0.92 0.31Hdac1 Histone deacetylase 1 0.90 0.29 0.97 0.40 0.98 0.45Hdac2 Histone deacetylase 2 0.98 0.46 1.09 0.26 1.12 0.12Hdac3 Histone deacetylase 3 0.90 0.31 1.26* 0.04 1.10 0.25Mbd3 Methyl-CpG binding domain protein 3 0.74** 0.01 0.94 0.31 0.91 0.33Mecp2 Methyl CpG binding protein 2 0.85 0.17 1.08 0.33 0.78 0.08Parp1 Poly (ADP ribose) polymerase family, member 1 1.25* 0.03 0.99 0.48 0.95 0.35Rnf4 RING finger protein 4 1.07 0.20 1.11 0.28 0.86 0.13

Asterisks indicate fold change that was significantly different, as determined by Student t tests: *P < 0.05, **P < 0.01.

Jia et al. PNAS | Published online December 22, 2014 | E57

NEU

ROSC

IENCE

PNASPL

US

with normal STHdhQ7 striatal cells (26). Also consistent with theseprevious studies was that a large majority of the DNA methylationchanges occurred in CpG-poor regions of the genome, comparedwith CpG-rich regions (Fig. S1). Using the program GenomicRegions Enrichment of Annotations Tool (GREAT), whichassociates input genomic regions with their putative target genesand then assigns significant functional annotations (27), we foundthat genes associated with methylation changes at either CpG-poor or CpG-rich regions were enriched for developmental pro-cesses, cell differentiation, gene transcription, chromatin binding,and immune signaling (Table S3), similar to previous reports (26).We next identified methylated positions that were regulated by

HDACi 4b exposure in human fibroblast cells from both normalcontrols and patients with HD. All differentially methylatedpositions (DMPs) were associated with one or two annotatedgenes, and 87% of the DMPs were within 500 kb of an annotatedtranscription start site. A greater proportion of genomic sites werefound to be decreased in methylation by HDACi 4b treatment inHD cells compared with normal cells (62% and 48% for HD andnormal cells, respectively) (Fig. 1). GREAT analysis found thatHDACi 4b-induced DMPs were significantly associated with thebiological processes of spermatogenesis, the multicellular organis-mal reproductive process, and antigen presentation and processing(Table 2). In addition, both sets returned a significant relationshipto the human phenotype ontology enrichment category of Y-linked

inheritance and gonosomal inheritance (Table 2). From the humandataset of DMPs, 35.2% and 42.3% of the human genomic coor-dinates (hg19) in the WT and HD cells, respectively, had a corre-sponding region in the mouse genome (mm10). Among thedifferentially methylated genes that showed a corresponding lift-over to mouse were Kdm5d (also known as Jarid1d or Smcy), sex-determining region of Chr Y (Sry), eukaryotic translation initiationfactor 1A, Y-linked (Eif1ay), and synaptoporin (Synpr). HDACi4b did not alter the expression of any known imprinted genes,as determined by cross-referencing the Geneimprint database(geneimprint.com/).We focused on one of these Y-linked genes, Kdm5d, further

testing whether HDACi 4b could alter its methylation status inmouse sperm DNA from N171-82Q transgenic mice. Aftertreatment of male mice for 1 mo with HDACi 4b, using meth-ylated DNA immunoprecipitation in combination with real-timequantitative PCR (qPCR), we demonstrated increases in meth-ylation elicited by HDACi 4b treatment at three different Kdm5dloci in DNA from mouse sperm (Fig. 2).DNA methylation is an epigenetic modification that can be

inherited in the germ line and can lead to transgenerational effectson subsequent progeny. Therefore, we compared behavioral phe-notypes in offspring from HDACi 4b- and vehicle-treated maleHD mice between 14 and 16 wk of age. We found that first filialgeneration (F1) transgenic offspring from HDACi 4b-treated male

WT cells

WT1

-Veh

WT2

-Veh

WT3

-Veh

WT4

-4b

WT5

-4b

WT6

-4b

Pro

be ID

Chr

. P

ositi

on

RE

FGE

NE

ID

52 D

MP

s

KDM5D

SRY

TTTY14

EIF1AY

HD cells

HD

1-V

eh

HD

2-V

eh

HD

3-V

eh

HD

4-4b

HD

5-4b

HD

6-4b

50 D

MP

s

SYNPR

TSPY4

KDM5D

EIF1AY

Pro

be ID

Chr

. P

ositi

on

RE

FGE

NE

ID

Fig. 1. Differential DNA methylation due to HDAC inhibitor treatment. Data for CpG sites differentially methylated between HDACi 4b-treated and vehicle-treated (delta beta > 0.2) in normal and HD human fibroblasts on the 27K DNA methylation array. Yellow denotes a decrease in methylation due to HDACi 4btreatment; black denotes an increase in methylation. The total number of DMPs in each cell type is shown along the y axis. Chr., chromosome, SRY, sex-determining region of Chr Y; SYNPR, synaptoporin; TSPY4, testis-specific protein, Y-linked 1; TTTY14, testis-specific transcript, Y-linked 14.

E58 | www.pnas.org/cgi/doi/10.1073/pnas.1415195112 Jia et al.

transgenic F0 mice show significantly improved motor function inthe rotarod test, improved cognition in the alternating T-maze test,and increased motor function in the open-field activity test (Fig. 3B–E), compared with F1 transgenic offspring from vehicle-treatedmale transgenic F0 mice. Interestingly, these improvements weremore pronounced in male transgenic mice compared with femaletransgenic mice (Fig. 3 C–E). No difference in the body weight of

either sex was observed (Fig. 3A). No differences in CAG repeatlength were observed in HDACi 4b-exposed offspring comparedwith their parents (82.4 ± 1.1 vs. 82.0 ± 1 CAG repeats). Wefurther tested for transgenerational effects of another inhibitor,RGFP966 (Repligen Corporation), that is structurally related toHDACi 4b but selectively targets the HDAC3 subtype (6).RGFP966 treatment beginning at 8 wk of age showed direct

Table 2. GREAT analysis of HDACi 4b-induced DNA methylation differences in human HD fibroblasts

Term Binomial P value Binomial FDR Q value Binomial FE

Top enriched GO biological process termsSpermatogenesis 1.43E-11 3.14E-08 8.84889Sexual reproduction 2.29E-10 4.02E-07 6.123311Gamete generation 2.56E-09 2.49E-06 6.19334Gonadal mesoderm development 6.54E-09 5.73E-06 79.2003Multicellular organismal reproductive process 2.29E-07 1.18E-04 4.483269Antigen presentation via MHC class II 2.53E-07 1.23E-04 77.53082Reproductive process 1.37E-06 5.46E-04 3.301294Reproduction 1.43E-06 5.43E-04 3.292752Antigen processing and presentation 1.22E-05 3.81E-03 16.96118Mesoderm development 7.11E-04 1.00E-01 7.069603Sex differentiation 3.28E-03 3.55E-01 3.567114Gonad development 4.51E-03 4.44E-01 3.853899Development of primary sexual characteristics 8.14E-03 6.99E-01 3.402624Reproductive structure development 9.96E-03 8.16E-01 3.257634Developmental process involved in reproduction 1.80E-02 1.00E+00 2.577171Immune system process 8.69E-02 1.00E+00 1.6056

Top enriched human phenotype termsY-linked inheritance 2.90E-19 1.74E-15 143.0223Gonosomal inheritance 4.64E-08 1.39E-04 10.1296

FDR, false discovery rate; FE, fold enrichment.

Fold

-enr

ichm

ent i

nD

NA

met

hyla

tion

H19 Iap

Gapdh

0

2

4

6

8 (+) control (+) control (-) controlKdm5d - upstream

Fold

-enr

ichm

ent i

nD

NA

met

hyla

tion

Vehicl

e

HDACi 4b

Vehicl

e

HDACi 4b

0

1

2

3

4

Site 1 Site 2

*

Kdm5d - downstream

Vehicl

e

HDACi 4b

Vehicl

e

HDACi 4b

0

2

4

6

Site 3 Site 4

**

A

D Kdm5d locus:

CB

Fig. 2. Effect of HDACi 4b on DNA methylation at the Kdm5d locus in mouse sperm DNA. The DNA methylation array revealed two regions increased inmethylation in the Kdm5d locus with HDACi 4b treatment. Two pairs of primers were designed to verify each region in ±100 bp of the methyl-array probesite. Sites 1 and 2 are designed to detect a promoter region of 250 bp upstream of the Kdm5d transcription start site. Sites 3 and 4 are designed to detecta region 30.18 kb downstream of the Kdm5d transcription start site. (A) HDACi 4b treatment significantly increased the methylation signals in the promoter“site 1” region of Kdm5d but not in the “site 2” region. (B) Both “sites 3 and 4” increased the methylation signal with HDACi 4b treatment. (C) Promoter CpGregion of glyceraldehyde-3-phosphate dehydrogenase (Gapdh) is a negative control, and Intracisternal A-particle (IAP) and H19 imprinted maternallyexpressed transcript (H19) are maternally imprinted positive controls for DNA methylation. All relative methylation was normalized to the methylation levelof the Gapdh promoter region. Statistical differences were determined by a two-tailed Student t test: *P < 0.05 vs. vehicle group (n = 6).

Jia et al. PNAS | Published online December 22, 2014 | E59

NEU

ROSC

IENCE

PNASPL

US

beneficial effects on HD disease phenotypes in N171-82Qtransgenic mice (F0 generation measured at 12–16 wk of age)(Fig. S2). These results are highly similar to our previous studiesusing HDACi 4b (9). However, we did not observe any signifi-cant transgenerational effects of RGFP966 on the F1 generationof N171-82Q transgenic mice (i.e., the offspring of RGFP966-treated fathers) (Fig. S3).We next tested Kdm5d expression levels in these same off-

spring at 16 wk of age. F1 transgenic offspring from HDACi 4b-treated male transgenic mice showed a 7.8-fold increase in thecortical expression of Kdm5d compared with F1 transgenic off-spring from vehicle-treated male transgenic mice (Fig. 4A).Additional genes from our DNA methylome analysis were testedfor gene expression difference in the cortex of male offspring, andonly Eif1ay showed a small but significant difference (Fig. S4).Consistent with the Kdm5d mRNA findings, we detected an in-crease in Kdm5d protein expression in the cortex of F1 transgenicoffspring from HDACi 4b-treated males by Western blot analysis(Fig. 4B). Because Kdm5d is a Lys-specific demethylase, whichspecifically acts at H3K4, we also measured histone methylation atH3K4, and found lower levels in the brains of F1 transgenic off-spring from HDACi 4b-treated males vs. F1 transgenic offspringfrom vehicle-treated males (Fig. 4B). We found that Kdm5d ex-pression levels were negatively correlated with H3K4 methylation(H3K4me) levels (Fig. 4C), as would be predicted.

We further tested the direct role of elevating Kdm5d expres-sion in the mechanism for improving HD phenotypes usingconditionally immortalized homozygous mutant STHdhQ111

striatal neuronal progenitor cells. These cells express full-lengthendogenous huntingtin (Htt) protein containing 111 glutaminesand display a phenotype of decreased mitochondrial functionand energy metabolism compared with WT cells expressing Httwith seven glutamines (28, 29). Overexpression of Kdm5d inSTHdhQ111 striatal cells resulted in a significant improvementin the Htt-induced metabolic deficit (Fig. 4D).

DiscussionIn summary, we demonstrated that HDAC1/3 inhibition canelicit changes in the expression of genes related to DNA meth-ylation in mouse brain and cause selected alterations in CpGmethylation in DNA from control and HD human fibroblasts.Methylation of CpG sites induced by HDACi 4b was significantlyenriched at Y chromosome genes, including EIF1AY, TSPY4,and KDM5D. KDM5D/Kdm5d showed increased methylation atseveral CpG loci not only in human cells, but also in mousesperm DNA from HDACi 4b-treated male mice. HD transgenicF1 offspring from drug-treated male HD transgenic miceexhibited improved disease phenotypes compared with trans-genic offspring from vehicle-treated HD transgenic mice.The behavioral differences were sexually dimorphic, with males

0 4 8 12 160

100

200

300

400

Trials (4 trials per day; 4 days)

FemalesF1HDTg-Veh F1HDTg-4b

0 4 8 12 160

100

200

300

Trials (4 trials per day; 4 days)

Late

ncy

tofa

ll (s

ec)

Males

***, P<0.0001;F(1,368)=68.5

F1HDTg-Veh F1HDTg-4b

Body weight T-maze

Open field activity - Males: Open field activity - Females:

Rotarod

F1WT-V

eh

F1WT-4b

F1HDTg-V

eh

F1HDTg-4b

2

4

6

8

10***

*

N.S.

Cor

rect

choi

ces

in10

tria

ls

Age (weeks)

Bod

yw

eigh

t(g)

6 8 10 12 14 1610

15

20

25

30 Male F1HDTg-VehMale F1HDTg-4bFem F1HDTg-VehFem F1HDTg-4b

0 2 4 6 8 1050

100

150

200

250F1HDTg-Veh F1HDTg-4b

Time (min)

Dis

tanc

e(U

nits

)

**, P=0.009; F(1,171)=6.9

0 2 4 6 8 102

4

6

8

Time (min)

Ambu

lato

ryTi

me

(sec

)

*, P=0.047; F(1,173)=3.9

0 2 4 6 8 108

10

12

14

16

Time (min)

Ster

eoty

pic

time

(sec

)

*, P=0.01; F(1,150)=6.1

0 2 4 6 8 1010

20

30

40

50

60

Time (min)

Aver

age

Velo

city

*, P=0.02; F(1,186)=5.1

0 2 4 6 8 100

50

100

150

200

250F1HDTg-Veh F1HDTg-4b

Time (min)

Dis

tanc

e(U

nits

)

0 2 4 6 8 100

2

4

6

8

Time (min)

Ambu

lato

ryTi

me

(sec

)

0 2 4 6 8 108

10

12

14

16

Time (min)

Ster

eoty

pic

time

(sec

)

0 2 4 6 8 100

20

40

60

80

Time (min)

Aver

age

Velo

city

A

D E

B C

Fig. 3. Comparisons of disease phenotypes in offspring of male N171-82Q transgenic mice treated with HDACi 4b (F1HDTg-4b) relative to offspring of N171-82Q transgenic mice treated with vehicle (F1HDTg-veh). (A) Effects of HDACi 4b exposure on the body weights of N171-82Q transgenic mice. (B) Rotarodperformance of offspring of vehicle- and HDACi 4b-exposed N171-82Q transgenic mice. Males and females are shown separately. Two-way ANOVA revealedsignificant differences between male F1HDTg-4b mice compared with male F1HDTg-veh mice, whereas female mice did not show statistically significantdifferences. N.S., not significant. (C) Number of correct choices of first 10 trials in the alternating T-maze test. One-way ANOVA revealed significant dif-ferences between vehicle-exposed WT and N171-82Q transgenic mice (**P < 0.001) and a significant difference between vehicle-exposed and HDACi 4b-exposed N171-82Q mice (*P < 0.05). Both male and female mice were combined in this assay. (D and E) Effects of HDACi 4b exposure on open-field activity ofN171-82Q transgenic mice over a 10-min test period. Two-way ANOVA revealed significant differences between HDACi 4b-exposed and vehicle-exposedN171-82Q transgenic mice for distance, ambulatory time, stereotypic time, and average velocity only in male mice, as indicated.

E60 | www.pnas.org/cgi/doi/10.1073/pnas.1415195112 Jia et al.

showing more pronounced improvements compared with females;this effect is consistent with the high proportion of Y chro-mosome-linked genes showing differential methylation in re-sponse to HDACi 4b.Elevation in the expression of Dnmt3a, which establishes new

DNA methylation patterns, or Dnmt1, a maintenance DNAmethyltransferase, could account for increases in DNA methyl-ation, whereas elevated expression of genes encoding putativeDNA demethylase enzymes, such as Gadd45b and Parp1, wouldlead to decreases in DNA methylation. From our global meth-ylation profiling of human fibroblasts, we found that HDACi 4belicited both increases and decreases in CpG methylation, al-though a greater number of sites were found to be decreased inmethylation in HD cells compared with WT cells. These effectsmight explain, in part, why many microarray studies have revealedthat HDAC inhibitors can cause both increases and decreases ingene expression (11, 15).The effect of HDAC inhibitors to alter DNA methylation is

highly relevant because DNA methylation is an epigenetic mo-dification that can be inherited in the germ line and can lead totransgenerational effects on subsequent progeny. The ability ofenvironmental factors to promote a phenotype or disease statenot only in the individual exposed but also in offspring is termed“epigenetic transgenerational inheritance.” Epigenetic transgen-erational inheritance of altered phenotypes has been observed inmany species, such as worms, flies, plants, rodents, and humans(30–34), suggesting that ancestral environmental exposure can

influence disease epidemiology. Importantly, studies have dem-onstrated transgenerational effects via paternal transmission (35,36), even in humans (33, 34), and these effects can be detected inthe F1 generation (compared with maternal transmission, whichrequires testing of the F2 and F3 generations). This effect mostlikely involves alterations in the sperm DNA methylome. Afterfertilization, the paternal and maternal alleles are demethylated,in part, to develop the pluripotent state of the ES cells, althoughit is now known that methylation clearing is not complete afterfertilization (37–39). We propose that HDAC inhibitors can in-duce germ-line epigenetic modifications and become permanentlyprogrammed similar to the DNA methylation of an imprintedgene (38, 39). Accordingly, we show increased methylation atseveral CpG loci of the Kdm5d gene in mouse sperm DNA fromHDACi 4b-treated male mice.The transgenerational effects elicited by exposure to HDACi

4b were not mimicked by a related HDAC inhibitor, RGFP966,which selectively targets the HDAC3 subtype. HDACi 4b po-tently inhibits HDAC3 and HDAC1, but it also inhibits HDAC2and HDAC8 at higher concentrations (micromolar) (10). Hence,inhibition of other class I HDACs, particularly HDAC1, could beresponsible for the DNA methylation changes we observed, aswell as in the ensuing downstream effects.One important candidate gene we have implicated in this study

is KDM5D/Kdm5d, which encodes a histone Lys (K)-specificdemethylase specific for H3K4. We found that Kdm5d expressionwas elevated in the brains of F1 offspring of HDACi 4b-exposed

H3K4me

Total H3

F1HDTg-veh F1HDTg-4b

Kdm5d

Gapdh

F1HDTg-veh F1HDTg-4b

A

C

B

Amount of Kdm5d cDNAMock

0.5 ug

1 ug

2 ug

0.0

0.2

0.4

0.6

0.8

*** ***

HD-Q111 cells

*

Met

abol

ic Im

prov

emen

t(A

bsor

banc

e)

F1WT-ve

h

F1WT-4b

F1HDTg-ve

h

F1HDTg-4b

0

10

20

30

40

50

Gen

e ex

pres

sion

(Hpr

t nor

mal

ized

)***

**

D

Fig. 4. (A) Real-time qPCR results showing altered expression of Kdm5d in cortex of male offspring of vehicle-treated or HDACi 4b-treated male mice. Valuesshown are the mean ± SEM expression value from n = 4–5 mice per group. One-way ANOVA was used to determine significant differences, and these areindicated by asterisks: **P < 0.001; ***P < 0.0001. (B) Western blot analysis showing elevated Kdm5d protein and reduced H3K4me expression in cortex ofmale offspring of vehicle-treated (F1HD-veh) and HDACi 4b-treated (F1HD-4b) male HD transgenic mice. Representative blots are shown, although n = 8F1HD-veh and n = 9 F1HD-4b samples were analyzed in total. (Right) Bar graph quantification of band signals is shown. Student t test was used to determinesignificance, as indicated by asterisks: *P < 0.05. (C) Correlation between Kdm5d expression levels and H3K4me levels. Pearson correlation analysis of Westernblot data for Kdm5d and H3K4me was used, with each point representing a different sample (R = −0.502, P = 0.045). (D) Improvement of the Htt-elicitedmetabolic deficits in STHdhQ111 striatal cells elicited by KDM5D overexpression. Cells were transfected with human KDM5D cDNA using FuGENE transfectionreagent and XTT assays performed after 24 h. Significant differences were determined by one-way ANOVA, followed by a Dunnett’s posttest comparing tothe mock-transfected control: *P < 0.05; ***P < 0.0001.

Jia et al. PNAS | Published online December 22, 2014 | E61

NEU

ROSC

IENCE

PNASPL

US

fathers, in correlation with reduced levels of H3K4me. H3K4meis an important epigenetic mark that has been implicated in di-verse biological functions, such as gene activation and repression,the DNA damage response, and genomic imprinting and devel-opment (40). Proteins that regulate the methylation of H3K4 areknown to be involved in neurological and developmental dis-orders, and mutations in genes encoding histone demethylaseenzymes have been identified in autism, Rett syndrome, and X-linked mental retardation (XLMR) (41). For example, mutationsin the gene KDM5C, a closely related family member to KDM5Dbut expressed on the X chromosome (40), are associated withXLMR and result in reduced demethylase activity, which is det-rimental to neuronal function (41). Microarray studies have shownthat Kdm5c expression is decreased in human HD caudate (42)and in the striatum of R6/2 mice (43), which is consistent with ourcurrent studies on Kdm5d in N171-82Q mice, implicating reduceddemethylase activity in HD pathology. It should be noted thatanother study reported an increase in KDM5C expression in hu-man HD caudate (44); however, the discrepancy could be due tothe X-linked expression of KDM5C and the unknown sexes ofthese patients in that study.Of particular relevance to our work is a recent study impli-

cating a related H3K4 dimethyl demethylase, Spr-5, in trans-generational epigenetic effects observed in Caenorhabditis elegans(41). These researchers showed that Spr-5 mutations resulted inreduced fertility in C. elegans across successive generations (41).These findings further support a role for the mammalian H3K4demethylase, Kdm5d, in the transgenerational effects we observedin this study.Previous work studying epigenetic transgenerational inher-

itance has shown that diet, endocrine disruptors, and toxic com-pounds can lead to heritable epigenetic alterations of the germline, and can subsequently significantly affect the transcriptomesand phenotypes of following generations (42, 43). Typically, theresulting phenotypic effects are unwanted and/or detrimental (i.e.,infertility or toxic effects in the offspring). In stark contrast, ourresults show that HDAC inhibitors can impart beneficial effects tooffspring in a transgenerational manner. These findings will havesignificant implications for human health because they enforce theconcept that ancestral drug exposure may be a major molecularfactor that can affect disease phenotypes, yet in a positive manner.This phenomenon is especially relevant for HDAC inhibitors,which are increasingly being used to treat a broad range of humandiseases.

Materials and MethodsMice and Drug Treatments. An B6C3-Tg(HD82Gln)81Dbo/J (N171-HD82Q) line(Jackson Laboratories) has been maintained at The Scripps Research Instituteby breeding male heterozygous N171-HD82Q mice with F1 hybrids of thesame background. The CAG repeat lengths in these mice were verified bycommercial genotyping (Laragen) and found to be 82 ± 1 CAGs. The lifespanof the N171-82Q HD mouse is ∼17–20 wk, with HD-like symptoms beginningat 10–12 wk of age. HDACi 4b [N1-(2-aminophenyl)-N7-phenylheptanediamide]was synthesized by Scynexis Corporaton as described previously (7). N171-82Q transgenic mice were housed and maintained on a normal 12-h light/dark cycle with lights on at 6:00 AM and free access to food and water. Formicroarray and gene expression studies, groups of male mice (n = 6–8) wereadministered HDACi 4b (50 mg/kg) or vehicle for 10–12 wk by s.c. injection(three injections per week) beginning at 8 wk of age. For sperm DNAmethylation studies and breeding, groups of male mice (n = 6–8) were ad-ministered HDACi 4b (50 mg/kg) or vehicle for 4 wk by s.c. injection (threeinjections per week) beginning at 8 wk of age. In separate studies, RGFP966was administered in a similar manner (SI Materials and Methods). HDACi 4bwas dissolved with 75% (vol/vol) polyethylene glycol 200; control mice re-ceived double-distilled H2O containing an equal volume of drug vehicle. Ourprevious studies have found that the s.c. injection paradigm for HDACi 4b isoptimal for improving disease phenotypes in N171-82Q transgenic mice,given the poor solubility of HDACi 4b for drinking water paradigms (44, 45).

Bodyweightswere recorded two times perweek.Micewere killed 6 h afterthe final injection. Brain regions (striatum and cortex) and gastrocnemius and

soleus skeletal muscle samples were dissected out for gene expression assays.All procedures were in strict accordance with the NIH Guide for the Care andUse of Laboratory Animals (46). All experimental protocols were approvedby the Animal Care and Use Committee at the Scripps Research Institute.

Cell Culture of Human Fibroblasts. HD patient and control fibroblast cells werepurchased from the Coriell Institute. HD and control fibroblast cells werecultured to 80% confluence in DMEM with 10% (wt/vol) FBS and 1% anti-biotic penicillin-streptomycin according to the instructions of the supplier.Cells were plated in 12-well plates, treated with HDACi 4b at 500 nM for 48 h,and then harvested to extract genomic DNA (gDNA) using a QIAamp DNAMini Kit (Qiagen) for methyl-array analysis.

Microarray Gene Expression Analysis. RNA was extracted from gastrocnemiusmuscle samples dissected from vehicle- and drug-treated mice and convertedto cDNA by standard protocols. RNA (200 ng) was amplified, biotinylated, andhybridized to Illumina MouseRef-8 v2.0 Expression BeadChips per the man-ufacturer’s protocol. Slides were scanned using Illumina BeadStation, andsignals were extracted using Illumina BeadStudio software. Gene expressionprofiles were normalized using robust multiarray average and log2-trans-formed using the Bioconductor package “beadarray.” Differential expres-sion was performed using the Bioconductor package “Limma” and adjustedP values for multiple testing, as described previously (43). The full microarraydataset can be found on the NCBI Gene Expression Omnibus (GEO) website,under accession no. GSE56963.

Functional Pathways Analysis. The functional relevance of microarray datagenerated from different brain regions from HDACi 4b-treated HDmice fromour previous studies (GEO accession no. GSE26317) and from muscle samplesin the current study (GEOaccessionno.GSE56963)were analyzedusing IngenuityPathways Analysis (Ingenuity Systems, www.ingenuity.com) and DAVID.

Real-Time qPCR Analysis. The real-time qPCR experiments were performedusing the ABI StepOne Detection System (Applied Biosystems) as describedpreviously (43). Specific primers for each sequence were designed usingPrimer Express 1.5 software (Applied Biosystems), and their specificity forbinding to the desired sequences was searched against the National Centerfor Biotechnology Information database (Tables S4 and S5). The amount ofcDNA in each sample was calculated using SDS2.1 software (Applied Bio-systems) by the comparative threshold cycle (Ct) method and expressed as2exp(Ct) using hypoxanthine guanine phosphoribosyl transferase (Hprt) asan internal control. Statistical significance was determined using Student ttests (unpaired and one- or two-tailed). All statistical tests were performedusing GraphPad software (GraphPad Prism).

DNA Methylation Profiling. DNA from normal and HD human fibroblasts wasquality controlled (DNA1000 Kit and BioAnalyzer 2100; Zymo Research) andbisulfate-converted (EZ DNA Methylation Kit; Zymo Research) according toeach manufacturer’s protocol. Bisulfite-converted DNA was hybridized toInfinium HumanMethylation27K BeadChips (Illumina, Inc.), scanned with aniScan (Illumina, Inc), and quality-controlled using GenomeStudio. For 27Kdata, β values for each probe were range-scaled with data collected fromDNA controls that were fully methylated [DNA methyltransferase-treated(New England Biolabs), bisulfite-converted DNA], unmethylated (untreatedgDNA), and half-methylated (50:50 mix of methylated and unmethylatedcontrols). For HD vs. normal comparisons, we used a multiple hypothesiscorrected P value of 0.001 and a difference in methylation cutoff of betachange of >j0.2j. Due to the lower number of methylation changes inducedby HDACi 4b treatment, a statistical cutoff of P < 0.1 and a beta change ofj0.2j were used.

Methylated DNA Immunoprecipitation Real-Time qPCR. Male transgenic N171-HD82Qmicewere treated daily with HDACi 4b (50mg/kg) for 2wk. The spermwas released from the cauda epididymis, and the gDNA was extracted andpurified using a Qiagen genome DNA purification kit as described previously(47). The purified gDNA was immunoprecipitated and subjected to real-timeqPCR analysis according to previous protocols (48). Briefly, gDNA was soni-cated with an S220 Focused-Ultrasonicator (Covaris, Inc.) and verified for anaverage size of 250–350 bp. Two micrograms of sonicated gDNA wasimmunoprecipitated with 1 μg of anti-5mC antibody (catalog no. BI-MECY-0100; Eurogentec) by incubation for 2 h at 4 °C with shaking. Sheep-M280anti-mouse Dynabeads (Life Technologies) were incubated with the samplefor another 2 h at 4 °C with shaking, followed by collection of the Dynabeadsusing a magnetic rack. The DNA binding magnetic beads were resuspended in

E62 | www.pnas.org/cgi/doi/10.1073/pnas.1415195112 Jia et al.

proteinase K digestion buffer and incubated for 3 h at 50 °C to release gDNA.The immunoprecipitated gDNA was purified using a Qiagen MiniElute PCRPurification Kit followed by qPCR analysis to verify the target region.

Motor Behavioral Assessments of N171-82Q Offspring. In total, n = 55 F1 off-spring from F0 vehicle-treated male mice and n = 54 F1 offspring from F0HDACi 4b-treated male mice (50-mg/kg dose) from >25 different parentalmatings were tested.Rotarod. An AccuRotor rotarod (AccuScan Instruments) was used to measuremotor coordination and balance. Mice were tested during the light phase ofthe 12-h light/dark cycle by using an accelerating rotation paradigm. Micewere placed on a rotarod accelerating from 0 to 20 rpm over 600 s, andlatency to fall was recorded.Open-field test. Open field exploration was measured in a square Plexiglaschamber (27.3 cm × 27.3 cm × 20.3 cm; Med Associates, Inc.). Several be-havioral parameters (ambulatory distance, ambulatory time, stereotypictime, vertical time, vertical counts, and mean velocity) were recorded duringa 10-min observation period.Alternating T-maze. The alternating T-maze test was performed using methodssimilar to those described previously (49). The T-maze is made of transparentPlexiglas with a central arm (75 cm long × 12 cm wide × 20 cm high) and twolateral arms (32 cm long × 12 cm wide × 20 cm high) positioned at a 90°angle relative to the central arm. Forced alternation was used for the T-mazetraining. For T-maze testing, mice were provided an initial free choice for ei-ther arm of the maze, and the percentage of alternation over the next ninetrials was determined. This percentage can be used as an index of workingmemory performance (50). Statistical analyses for all behavioral tests wereperformed using one-way or two-way ANOVA (GraphPad Prism).

Striatal Cell Culture. Conditionally immortalizedWT STHdhQ7striatal neuronalprogenitor cells expressing endogenous normal Htt, with seven glutamines,

and homozygous mutant STHdhQ111 striatal neuronal progenitor cell linesexpressing endogenous mutant Htt, with 111 glutamines, were a kind do-nation from Marcy MacDonald, Massachusetts General Hospital, Boston, andgrown as described previously (51). Cells were plated in 24- or 96-well tissueculture plates. The following day, the cells were transfected with an ex-pression plasmid containing the cDNA of human KDM5D (NCBI Nucleotidedatabase, accession no. BM455812), which was directionally cloned into theexpression vector pCMV-Sport6. Transfection was carried out using FuGENEtransfection reagent (Promega) according to the manufacturer’s instruc-tions. Empty expression vector, pCMV, was used as a control. Metabolicassays were performed 24 h after shifting the cells to 39 °C by adding theXTT [2,3-Bis-(2-Methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5-carboxanilide,disodium salt] reagent (Sigma–Aldrich), followed by absorbance readings at490 nm. Successful transfection of KDM5D was assessed by measuring over-expression of Kdm5d protein usingWestern blotting with a specific anti-Kdm5dantibody (1:1,000 dilution; Acris). Significant differences in metabolic activitywere determined by one-way ANOVA, followed by Dunnett’s posttest usingGraphPad software.

Western Blotting.Western blot assays were performed as described previously(9) using a specific anti-Kdm5d antibody (1:1,000 dilution; Acris) and anti-H3K4me3 (1:5,000 dilution; Millipore), and were normalized to anti-GAPDH(1:1,000 dilution; Abcam) or total histone H3 (1:8,000 dilution; Millipore).Chemiluminescent signals were visualized using Western blotting luminolreagent (Pierce), and bands were captured and quantified using FluorochemE image analyzer (Cell Biosciences).

ACKNOWLEDGMENTS. We thank Giovanni Coppola for performing themicroarray studies. These studies were funded by NIH Grant U01NS063953.

1. Cohen-Carmon D, Meshorer E (2012) Polyglutamine (polyQ) disorders: The chromatinconnection. Nucleus 3(5):433–441.

2. Deutsch SI, Rosse RB, Mastropaolo J, Long KD, Gaskins BL (2008) Epigenetic thera-

peutic strategies for the treatment of neuropsychiatric disorders: Ready for primetime? Clin Neuropharmacol 31(2):104–119.

3. Kazantsev AG, Thompson LM (2008) Therapeutic application of histone deacetylase

inhibitors for central nervous system disorders. Nat Rev Drug Discov 7(10):854–868.4. Cantley MD, Haynes DR (2013) Epigenetic regulation of inflammation: Progressing

from broad acting histone deacetylase (HDAC) inhibitors to targeting specific HDACs.Inflammopharmacology 21(4):301–307.

5. Grayson DR, Kundakovic M, Sharma RP (2010) Is there a future for histone deacetylase

inhibitors in the pharmacotherapy of psychiatric disorders? Mol Pharmacol 77(2):126–135.

6. Malvaez M, et al. (2013) HDAC3-selective inhibitor enhances extinction of cocaine-

seeking behavior in a persistent manner. Proc Natl Acad Sci USA 110(7):2647–2652.7. Herman D, et al. (2006) Histone deacetylase inhibitors reverse gene silencing in

Friedreich’s ataxia. Nat Chem Biol 2(10):551–558.8. Chou CJ, Herman D, Gottesfeld JM (2008) Pimelic diphenylamide 106 is a slow, tight-

binding inhibitor of class I histone deacetylases. J Biol Chem 283(51):35402–35409.9. Jia H, Kast RJ, Steffan JS, Thomas EA (2012) Selective histone deacetylase (HDAC)

inhibition imparts beneficial effects in Huntington’s disease mice: Implications for theubiquitin-proteasomal and autophagy systems. Hum Mol Genet 21(24):5280–5293.

10. Jia H, et al. (2012) Histone deacetylase (HDAC) inhibitors targeting HDAC3 andHDAC1 ameliorate polyglutamine-elicited phenotypes in model systems of Hunting-ton’s disease. Neurobiol Dis 46(2):351–361.

11. Thomas EA, et al. (2008) The HDAC inhibitor 4b ameliorates the disease phenotypeand transcriptional abnormalities in Huntington’s disease transgenic mice. Proc NatlAcad Sci USA 105(40):15564–15569.

12. Ferrante RJ, et al. (2003) Histone deacetylase inhibition by sodium butyrate chemo-therapy ameliorates the neurodegenerative phenotype in Huntington’s disease mice.

J Neurosci 23(28):9418–9427.13. Hockly E, et al. (2003) Suberoylanilide hydroxamic acid, a histone deacetylase in-

hibitor, ameliorates motor deficits in a mouse model of Huntington’s disease. Proc

Natl Acad Sci USA 100(4):2041–2046.14. Gardian G, et al. (2005) Neuroprotective effects of phenylbutyrate in the N171-82Q

transgenic mouse model of Huntington’s disease. J Biol Chem 280(1):556–563.15. Peart MJ, et al. (2005) Identification and functional significance of genes regulated by

structurally different histone deacetylase inhibitors. Proc Natl Acad Sci USA 102(10):

3697–3702.16. Glozak MA, Sengupta N, Zhang X, Seto E (2005) Acetylation and deacetylation of

non-histone proteins. Gene 363:15–23.17. Ribchester RR, et al. (2004) Progressive abnormalities in skeletal muscle and neuro-

muscular junctions of transgenic mice expressing the Huntington’s disease mutation.Eur J Neurosci 20(11):3092–3114.

18. Lodi R, et al. (2000) Abnormal in vivo skeletal muscle energy metabolism in Hun-tington’s disease and dentatorubropallidoluysian atrophy. Ann Neurol 48(1):72–76.

19. Arenas J, et al. (1998) Complex I defect in muscle from patients with Huntington’s

disease. Ann Neurol 43(3):397–400.20. Chaturvedi RK, et al. (2009) Impaired PGC-1alpha function in muscle in Huntington’s

disease. Hum Mol Genet 18(16):3048–3065.21. Barreto G, et al. (2007) Gadd45a promotes epigenetic gene activation by repair-me-

diated DNA demethylation. Nature 445(7128):671–675.22. Carey N, Marques CJ, Reik W (2011) DNA demethylases: A new epigenetic frontier in

drug discovery. Drug Discov Today 16(15-16):683–690.23. Hu XV, et al. (2010) Identification of RING finger protein 4 (RNF4) as a modulator of

DNA demethylation through a functional genomics screen. Proc Natl Acad Sci USA

107(34):15087–15092.24. Brown SE, Suderman MJ, Hallett M, Szyf M (2008) DNA demethylation induced by the

methyl-CpG-binding domain protein MBD3. Gene 420(2):99–106.25. Nakao M (2001) Epigenetics: Interaction of DNA methylation and chromatin. Gene

278(1-2):25–31.26. Ng CW, et al. (2013) Extensive changes in DNA methylation are associated with ex-

pression of mutant huntingtin. Proc Natl Acad Sci USA 110(6):2354–2359.27. McLean CY, et al. (2010) GREAT improves functional interpretation of cis-regulatory

regions. Nat Biotechnol 28(5):495–501.28. Cui L, et al. (2006) Transcriptional repression of PGC-1alpha by mutant huntingtin

leads to mitochondrial dysfunction and neurodegeneration. Cell 127(1):59–69.29. Lee JM, et al. (2007) Unbiased gene expression analysis implicates the huntingtin

polyglutamine tract in extra-mitochondrial energy metabolism. PLoS Genet 3(8):e135.30. Greer EL, et al. (2011) Transgenerational epigenetic inheritance of longevity in Cae-

norhabditis elegans. Nature 479(7373):365–371.31. Hauser MT, Aufsatz W, Jonak C, Luschnig C (2011) Transgenerational epigenetic in-

heritance in plants. Biochim Biophys Acta 1809(8):459–468.32. Ruden DM, Lu X (2008) Hsp90 affecting chromatin remodeling might explain trans-

generational epigenetic inheritance in Drosophila. Curr Genomics 9(7):500–508.33. Pembrey ME (2010) Male-line transgenerational responses in humans. Hum Fertil

(Camb) 13(4):268–271.34. Pembrey ME, et al.; ALSPAC Study Team (2006) Sex-specific, male-line transgenera-

tional responses in humans. Eur J Hum Genet 14(2):159–166.35. Carone BR, et al. (2010) Paternally induced transgenerational environmental re-

programming of metabolic gene expression in mammals. Cell 143(7):1084–1096.36. Ng SF, et al. (2010) Chronic high-fat diet in fathers programs β-cell dysfunction in

female rat offspring. Nature 467(7318):963–966.37. Hajkova P, et al. (2008) Chromatin dynamics during epigenetic reprogramming in the

mouse germ line. Nature 452(7189):877–881.38. Oswald J, et al. (2000) Active demethylation of the paternal genome in the mouse

zygote. Curr Biol 10(8):475–478.39. Morgan HD, Santos F, Green K, Dean W, Reik W (2005) Epigenetic reprogramming in

mammals. Hum Mol Genet 14(Spec No 1):R47–R58.40. Shao GB, et al. (2014) Dynamic patterns of histone H3 lysine 4 methyltransferases and

demethylases during mouse preimplantation development. In Vitro Cell Dev Biol

Anim 50(7):603–613.

Jia et al. PNAS | Published online December 22, 2014 | E63

NEU

ROSC

IENCE

PNASPL

US

41. Greer EL, et al. (2014) A histone methylation network regulates transgenerationalepigenetic memory in C. elegans. Cell Reports 7(1):113–126.

42. Junien C, Nathanielsz P (2007) Report on the IASO Stock Conference 2006: Early andlifelong environmental epigenomic programming of metabolic syndrome, obesityand type II diabetes. Obes Rev 8(6):487–502.

43. Skinner MK, Manikkam M, Guerrero-Bosagna C (2010) Epigenetic transgenerationalactions of environmental factors in disease etiology. Trends Endocrinol Metab 21(4):214–222.

44. Beconi M, et al. (2012) Oral administration of the pimelic diphenylamide HDAC in-hibitor HDACi 4b is unsuitable for chronic inhibition of HDAC activity in the CNS invivo. PLoS ONE 7(9):e44498.

45. Chen JY, et al. (2013) Effects of the Pimelic Diphenylamide Histone Deacetylase InhibitorHDACi 4b on the R6/2 and N171-82QMouseModels of Huntington’s Disease. PLoS Curr 5:5.

46. Committee on Care and Use of Laboratory Animals (1996) Guide for the Care and Useof Laboratory Animals (Natl Inst Health, Bethesda), DHHS Publ No (NIH) 85-23.

47. Chang HS, Anway MD, Rekow SS, Skinner MK (2005) Transgenerational epigenetic

imprinting of the male germline by endocrine disruptor exposure during gonadal sex

determination. Endocrinology 147(12):5524–5541.48. Weber M, et al. (2005) Chromosome-wide and promoter-specific analyses identify

sites of differential DNA methylation in normal and transformed human cells. Nat

Genet 37(8):853–862.49. Deacon RM, Rawlins JN (2006) T-maze alternation in the rodent. Nat Protoc 1(1):

7–12.50. Bontempi B, Whelan KT, Risbrough VB, Lloyd GK, Menzaghi F (2003) Cognitive en-

hancing properties and tolerability of cholinergic agents in mice: A comparative study

of nicotine, donepezil, and SIB-1553A, a subtype-selective ligand for nicotinic ace-

tylcholine receptors. Neuropsychopharmacology 28(7):1235–1246.51. Trettel F, et al. (2000) Dominant phenotypes produced by the HD mutation in STHdh

(Q111) striatal cells. Hum Mol Genet 9(19):2799–2809.

E64 | www.pnas.org/cgi/doi/10.1073/pnas.1415195112 Jia et al.